A single-objective EPR based model for creep index of soft clays considering L₂ regularization

A correlation of creep index (C_α) with high performance is highly recommended for soft clay engineering practice. Current empirical correlations are only acceptable for a few clays. This paper aims to propose a robust and effective evolutionary polynomial regression (EPR) model for C_α of clay. First, a database covering various clays is formed, in which 120 data are randomly selected for training and the remaining data are used for testing. To avoid overfitting, a novel EPR procedure using a newly enhanced differential evolution (DE) algorithm is proposed with two enhancements: (1) a new fitness function is proposed using the structural risk minimization (SRM) with L₂ regularization that penalizes polynomial complexity, and (2) an adaptive process for selecting the combination of involved variables and size of polynomial terms is incorporated. By comparing the predictive ability, model complexity, robustness and monotonicity, the EPR formulation for C_α involving clay content, plasticity index and void ratio with three terms is selected as the optimal model. A parametric study is then conducted to assess the importance of each input in the proposed model. All results demonstrate that the proposed model of C_α is simple, robust, and reliable for applications in engineering practice.